空间数据压缩的高光谱降维技术比较

发布时间：2018-07-06 19:37

本文选题：光谱数据降维 + 数据压缩　；参考：《遥感信息》2017年02期

【摘要】：针对从波段数目较多的海量高光谱遥感影像数据中高效地检索出所需信息这一迫切需要解决的问题,将空间数据压缩算法引入到光谱数据降维中,以规则树分组和曲线数据综合技术为基础,设计了十六叉树状变换、垂距光谱检索、偏角光谱检索、道格拉斯-普克光谱检索算子。采用相对光谱识别概率、相对光谱识别熵、相对光谱识别力3种统计学方法,从不同角度通过与常规的光谱角度制图和光谱信息熵进行比较分析发现:利用新的检索算子提取光谱曲线特征向量,进行相似性测度,降低了光谱检索的时间频率。在保证相近识别能力的条件下,能够大大提供高程序的检索效率,是几种快速有效的高光谱特征匹配和检索算子。
[Abstract]:In view of the urgent problem of efficiently retrieving the required information from the massive hyperspectral remote sensing image data with a large number of bands, the spatial data compression algorithm is introduced to reduce the dimension of spectral data. Based on the rule tree grouping and curve data synthesis techniques, the hexadecimal tree transform, vertical spectral retrieval, offset spectral retrieval and Dow Glass-Glask-spectral retrieval operator are designed. Using three statistical methods: relative spectral recognition probability, relative spectral recognition entropy and relative spectral recognition power. By comparing with the conventional spectral angle mapping and spectral information entropy, it is found that the new retrieval operator is used to extract the characteristic vector of the spectral curve and measure the similarity, thus reducing the time frequency of the spectral retrieval. Under the condition that the similar recognition ability is guaranteed, the retrieval efficiency of high program can be greatly provided, and it is a kind of fast and effective hyperspectral feature matching and retrieval operator.
【作者单位】：成都理工大学地学空间信息技术国土资源部重点实验室;成都理工大学地球科学学院;武汉大学学报信息科学版编辑部;
【基金】：国家自然科学基金(41071265、41102225、41201440) 高等学校博士学科点专项科研基金(201351221200092013) 成都理工大学骨干教师培养计划(DG0002)
【分类号】：TP751

【参考文献】